Introduction
Epidural fibrosis is one of the most common causes of failed back surgery syndrome (FBSS), and it is the causative factor of pain in up to 36% [1] of FBSS patients. Epidural fibrosis usually results from dural sac compression, nerve root tethering, or from interference with the nerve root vascular supply or cerebrospinal fluid flow [2]. Once the fibrosis forms, there is no effective treatment, and fibrous tissue excision surgery can also cause new epidural fibrosis [3]. One of the modalities for epidural fibrosis is the reduction or prevention of fibrosis formation. Many studies have developed formulations or techniques for its prevention. Most are animal model studies that test the efficacy of the drugs or materials against epidural fibrosis and then compare the study group and a control group based on the histopathology of epidural fibrosis. One of the common postoperative epidural fibrosis outcome measurements is the epidural fibrous adherence grading proposed by He et al. [4]. The classification definition proposed by He et al. [4], with additional definitions for the degree of severity, ranges from normal to a severe degree of epidural fibrosis, and is described as follows: grade 0 (normal, no epidural fibrosis), grade 1 (mild degree, thin fibrous band(s) over dura), grade 2 (moderate degree, continuous adherence observed but less than two-thirds of laminectomy defect) and grade 3 (severe fibrosis, scar tissue adherence large, more than two-thirds of laminectomy area, and/or extending to nerve roots). In the current animal model study, the postoperative epidural fibrous adherence classification was validated for its interrater reliability using percent agreement, kappa, and the intraclass correlation coefficient (ICC).
Materials and Methods
The Animal Research Ethics Committee of Khon Kaen University evaluated and approved the study protocol. Twenty female and 20 male, 8-week-old (adult) Sprague-Dawley rats were laminectomized at their L5-L6 spines. At six weeks after the operation, all of the rats were euthanized using a lethal dose of CO2. The L5-L6 vertebrae and all surrounding soft tissues were dissected en bloc. The specimens were fixed in phosphate buffered 10% formaldehyde for two days, decalcified in De Castro's fluid, dehydrated in alcohol, embedded in paraffin, and serially cut in a transverse fashion from L5-L6. Then, 5-µm cross-sections were processed for hematoxylin and eosin staining [5].
The histopathologic specimens with various degrees of epidural fibrosis were collected. Each of the epidural fibrosis histopathologic slides was labeled with a computer-generated randomized number, which were kept in sealed opaque envelopes. The slides with their number were then scanned with ScanScope FL (Aperio, Vista, CA, USA) into the digital files (ScanScope Virtual Slide). The five identical copies of these files were independently evaluated and graded for the degree of postoperative epidural fibrosis by the five assessors masked to the processes of the study.
1. Histopathologic grading of the post-laminectomy epidural fibrosis
As shown in Fig. 1, the dura and epidural area were assessed. The epidural fibrosis was graded at the area of the posterior aspect of the dura between the left and right rims of the laminectomy site (Fig. 2, point A to B). All assessors were oriented with the epidural fibrosis classification according Table 1 and Figs. 3 and 4. The extent of epidural fibrosis was rated as grade 0 (no fibrosis), grade 1 (mild degree), grade 2 (moderate degree), or grade 3 (severe degree) [4].
3. Statistical methods
The obtained results of histopathological grading were collected, and analyzed. The percent agreement, kappa, ICC and 95% confidence interval (CI) were calculated. All data were analyzed with STATA ver. 11.2 (StataCorp, College Station, TX, USA) and IBM SPSS ver. 20 (IBM Co., Armonk, NY, USA).
Results
Most (77.6%) of the distribution of epidural fibrosis grading by the assessors was between moderate and severe (Table 2). The percent agreement between the two assessors varied from 73.5% to 81.3%. The ICC ranged between 0.81 and 0.86 (Table 3).
The overall ICC of all 5 assessors was 0.83 (95% CI, 0.81-0.86). Grade 2 had a lower kappa (0.51) than the other grades, which were between 0.64 and 0.87 (Table 4).
Discussion
Due to the difficulty in obtaining human tissue for evaluation of postoperative epidural fibrosis, the animal model has an important role in such investigations. Another advantage of using an animal model is that the disease process (or pathological changes) is accelerated in animals, so it is possible to study the process in shorter time than in humans [7]. For post-laminectomy epidural fibrosis, the animal model used in our study was the Sprague-Dawley rat, which is relatively small but large enough to provide a suitable structure for a laminectomy and wound healing study.
Anatomical dissection of the rat's spine reveals an epidermis, basement membrane, sweat gland, dermis, and adipose tissue, similar to that found in humans [8]. The rat's vertebral endplate has an increasing diameter from the cervical to the lumbar spine-similar to humans, but unlike large quadrupeds-in which the spine diameter stays much the same over the whole length of the spine [9]. Consequently, we postulate that the epidural fibrosis seen in the rat model can be generalized to humans.
LaRocca and Macnab [10] demonstrated in a laminectomized dog model that peridural fibrosis could be identified in both the epidural space and the nerve roots when both are exposed. Nerve root fibrosis depended on the laminectomy procedure and whether it reached the nerve root. In the current study, only the epidural space was exposed after the laminectomy, so only the epidural area was evaluated for postoperative peridural fibrosis grading. Sen et al. [11] described that the epidural fibrosis tended to form a curvilinear pattern surrounding the dural sac following the contour of the inner laminar surface. In the laminectomy model, therefore, we rated the epidural fibrosis from the left to the right rim of the laminectomy site (Figs. 1,2,3).
According to the histopathological classification of postoperative epidural fibrosis in animal models, the classification by He et al. [4] is the most commonly used for prevention research [1213141516171819202122]. Most studies used a rat model, while some used rabbit and dog models. The classification is based upon the opinion of the pathologist who rates the degree of epidural fibrous adherence. Sucu et al. [23] found no validated histopathological grading system for epidural fibrosis. Therefore, the histopathological classification for postoperative epidural fibrosis experimental research should be validated and tested for its interrater reliability between the assessors who rate the postoperative epidural fibrosis. The kappa statistics or the chance agreement correction statistics are better than percentage agreement for describing interrater reliability [6]. Concerning the agreement statistics of the ordinal data, the weighted kappa or ICC [24] is preferable to Cohen's kappa [2526]. We used the quadratic-weighted kappa or ICC [24] because we considered partial agreement to be when the pathologists differently rated the epidural fibrosis for 1 or 2 grades and considered disagreement to be when the assessors differently rated the epidural fibrosis by more than 2 grades (e.g., one assessor rated grade 0 fibrosis while another rated grade 3 fibrosis). In the current study, therefore, we use the ICC for describing the interrater reliability instead of Cohen's kappa.
As mentioned above, the ICC (equivalent to the weighted kappa) [24] was suitable for reporting the interrater reliability of the epidural fibrosis classification, which was the ordinal parameter [27]. Weighted kappa credits the partial agreement between the raters. However, to test the reliability of each grade of epidural fibrosis, we adhered rigidly to the definition of each epidural fibrosis grade (not accepting partial agreement) and report the kappa (Table 4), instead of the weighted kappa or ICC.
According to the kappa statistics, Landis and Koch [28] suggested that a kappa (a) >0.75 (excellent agreement), (b) <0.40 (poor agreement), and (c) 0.40-0.75 (intermediate to good agreement). Viera and Garrett [29] further classified the interpretation of kappa (Table 5). The overall ICC of the epidural fibrous adherence in this study was 0.83 (p<0.001; 95% CI, 0.81-0.86). These represented the excellent interrater reliability of the classification for postoperative epidural fibrous adherence.
Grade 0 or 1 epidural fibrosis showed almost perfect agreement among the assessors, grade 2 indicated moderate agreement (kappa 0.51) and grade 3 had better agreement (Table 4). Grade 0 (no fibrosis), grade 1 (minimal fibrosis), and grade 3 (severe fibrosis) were not difficult to rate, while grade 2 (moderate fibrosis) was more difficult to rate since the ratings varied from mild to severe. We suggest that grade 2 should be rated with caution, and that consultation may be needed in equivocal cases.
We validated the epidural fibrosis classification based on histopathology in an animal model. For clinical relevance, Ross et al. [30] described the association between the amount of epidural fibrosis and recurrent radicular pain following lumbar discectomy. In their study, all patients underwent magnetic resonance imaging (MRI) at six months postoperatively. The amount of fibrosis was rated as no scar, mild, moderate, and severe scar. They reported that the patients who had severe epidural fibrosis were 3.2 times more likely to experience recurrent radicular symptoms than patients who had lesser degrees of fibrosis. Sen et al. [11] demonstrated in their clinical and MRI studies that the patients who had lesser degree of epidural fibrosis showed better clinical outcomes. We supposed that this epidural classification can be used in humans. However, because of the ethical issue and the difficulty in obtaining human tissue for the assessment of post-laminectomy epidural fibrosis, the classification definition can be applied to other investigational modalities, such as the axial image of the spinal MRI.
Conclusions
The histopathological classification of post-laminectomy epidural fibrosis showed excellent interrater reliability. The classification can be used for the assessment of epidural fibrosis and, certainly, for experimental animal model research for epidural fibrosis prevention or treatment.